very speed data processing system - Computer History...

34
a very hgh speed data processing system s5 k 0 'for Commerce, Industry, Science 'ENGLISH ELECTRIC' :.

Transcript of very speed data processing system - Computer History...

Page 1: very speed data processing system - Computer History Museumarchive.computerhistory.org/resources/text/English...a very hgh speed data processing system s5 k 0 'for Commerce, Industry,

a very hgh speed data processing system s 5

k 0 'for Commerce, Industry, Science

'ENGLISH ELECTRIC' :.I

Page 2: very speed data processing system - Computer History Museumarchive.computerhistory.org/resources/text/English...a very hgh speed data processing system s5 k 0 'for Commerce, Industry,
Page 3: very speed data processing system - Computer History Museumarchive.computerhistory.org/resources/text/English...a very hgh speed data processing system s5 k 0 'for Commerce, Industry,

The Lineage

On the pages that follow is described the 'English Electric ' KDF9 Data Processing system-one of the most advanced electronic digital systems yet devised.

KDF9 is the outcome of many years' experience accumulated by 'English Electric ' as designers, manufacturers, and users of computers, the Company being one of the pioneers in the field in Britain. In the early post-war years 'English Electric ' co-operated with the National Physical Laboratory in the development of the " ACE " Pilot Model, one of the first electronic digital computers made in the United Kingdom. The Company then proceeded with the development and production of a fully engineered computer which was to set new standards of speed and capacity. Great interest was shown in this undertaking and, literally at the drawing board stage, the Company was asked to build one each for the National Physical Laboratory and the Royal Aircraft Establishment, Farnborough. The new machine, which was named " DEUCE," was an immediate success and soon became widely known as the most powerful and advanced computer of its type. Successive marks were developed to meet the demand for increased speed and storage, all of which give the utmost satisfaction to their users.

As the use of computers became more widely known and appreciated there came increased demands for computers of much greater speed and capacity, and 'English Electric ' addressed themselves to the task of meeting these demands. A computer manufacturing division was established at Kidsgrove, North Staffordshire, and is now known as the Data Processing and Control Systems Division. From its inception the division has continuously expanded and produces

3 a wide variety of analogue and digital computers and ancillary equipment. These include the " LACE " series of analogue computers, multi-point scanners. "DATAPAC " systems of logical elements for process control and data handling, the KDPlO large-scale commercial data processing system, and now : the KDF9 general purpose computer for commerce, industry, and science.

Page 4: very speed data processing system - Computer History Museumarchive.computerhistory.org/resources/text/English...a very hgh speed data processing system s5 k 0 'for Commerce, Industry,

The success of 'English Electric' computers is in no small

measure due to the fact that the Company is one of the

largest USERS of analogue and digital computers and,

during the past decade, has accumulated a unique and intimate

knowledge of what an operator requires in a computer. Over , the years there has been built up a computer design team

which is second to none in forward thinking, backed by

a highly efficient manufacturing organisation, a prompt and

reliable after-sales service, and comprehensive user and

programming services.

Page 5: very speed data processing system - Computer History Museumarchive.computerhistory.org/resources/text/English...a very hgh speed data processing system s5 k 0 'for Commerce, Industry,

KDF9 DAT

STORAGE L

Main Intern; Nesting Stor Nesting Stor Nesting Store (DI

Q-Stores

PROGRAMMIh

Modular Instruct Main Store Trana,rlq Arithmetic Operation Peripheral Transfers Interlocks

CONTROL SYSTEMS Time-sharing Time-Smoothing Advance Control

OPERATING SPEED

Commercial Operations Computation Times

PERIPHERAL DEVICES

Magnetic Tape Units Paper Tape Punch High Speed Tape Reader High Speed On-line Printer Interruption Typewriter Punched Card Equipment

OFF-LINE OPERATIONS

High Speed Off-line Printer Xeronic Printer

USER SERVICES

TECHNICAL SYNOPSIS

Page 6: very speed data processing system - Computer History Museumarchive.computerhistory.org/resources/text/English...a very hgh speed data processing system s5 k 0 'for Commerce, Industry,

Key to Sys tern Must rat ion opposite

1. Power Supply Units (2) 2. Input/Output Units (2) 3. Main Store Units (3) 4. Main Control Units (3) 5. On-Line Magnetic Tape Units (2) 6. Off-Line Magnetic Tape Unit (1) 7. Printer Control Units (2) 8. Arithmetic Units (2) 9. On-Line Magnetic Tape Units (3)

10. Control Desk and Monitor Printer 11 . Paper Tape Punch 12. Paper Tape Reader 13. Off-Line Printer

Page 7: very speed data processing system - Computer History Museumarchive.computerhistory.org/resources/text/English...a very hgh speed data processing system s5 k 0 'for Commerce, Industry,
Page 8: very speed data processing system - Computer History Museumarchive.computerhistory.org/resources/text/English...a very hgh speed data processing system s5 k 0 'for Commerce, Industry,
Page 9: very speed data processing system - Computer History Museumarchive.computerhistory.org/resources/text/English...a very hgh speed data processing system s5 k 0 'for Commerce, Industry,
Page 10: very speed data processing system - Computer History Museumarchive.computerhistory.org/resources/text/English...a very hgh speed data processing system s5 k 0 'for Commerce, Industry,

The System

Designed and manufactured by the Data Processing and Control Systems Division of 'ENGLISH ELECTRIC ' the KDF9 all-British computing system embodies the wealth of experience of the Company as designers, manufacturers, and users of computer systems for commercial, industrial, and scientilic data processing applications. It is a general purpose system intended for general data processing applications and exceptionally fast large scale computation.

The keynote of the KDF9 design is USER CONVENIENCE and the utmost attention has been paid to the economics of its operation. An entirely new modular instruction code provides extreme simplicity of programming, and new standards of economy in the number of instructions and pro- gram storage space. In addition, the code facilitates translation from any problem language to an efficient machine program. The computing facilities include fixed or floating-point half-length working to increase effective storage and double-length working for problems demanding high precision.

KDF9 can be equipped with a very wide range of peripheral devices-including those associated with the complementary KDPlO system. There is no practical limit to the total number of devices which can be connected to the system, and any number can operate simultaneously, subject to a maximum theoretical instantaneous transfer rate of 1.33 million alphanumeric characters per second.

Transfers of blocks of data to or from peripheral units operate simultaneously with each other and with operations of the central computer. Up to four independent programs may operate con-currently by sharing the time of the control and arithmetic circuits. A time-smoothing advance control levels out peak demands on the main store and increases the already high speed of the computer by a factor of up to 35 %. The use of solid-state devices and printed circuits ensures maximum reliability, ease of maintenance has received careful attention, and all components are readily accessible. The external appearance, styling, and layout of KDF9 closely follow that of KDP10, and the durable and colourful finish of the cubicles continues the elegant standard established by the latter equipment.

A printed circuit "flip-fip" board

Page 11: very speed data processing system - Computer History Museumarchive.computerhistory.org/resources/text/English...a very hgh speed data processing system s5 k 0 'for Commerce, Industry,

r i l Storage Systems

MAIN INTERNAL STORE* KDF9 uses a ferrite matrix main store, transistor driven, with a

complete cycle time of 6 microseconds. One module of core store contains 4,096 words of 48 binary digits. These may be treated as 4,096 full-length words, 8,192 separately addressable half-length words, 32,768 6-bit alphanumeric characters, or from 8,192 to 24,576 KDF9 instructions.

The store is expansible and up to eight modules may be fitted to the computer, giving a total random access store capacity of 32,768 words. Provision is made in the instruction code for addressing even larger random access stores should this become necessary.

For scientific purposes any one word may be treated as a signed fixed-point number, or a floating-point number with 8-bit characteristic and 40-bit fraction. It may also be treated as a double-length or half-length fixed or floating-point number.

The diagram on page 10 clearly shows the various ways in which the KDF9 word may be used.

NESTING S T O W . KDF9 is the acme of advanced thought and design in electronic computers ; nowhere is this more true than in the design of the nesting store which combines maximum ecomony in programming with maximum operating speed.

A nesting store is, in effect, a series of sixteen registers holding a column of items of data, with the rule that an incoming item ioins the column at the tot, and causes all the items alreadv present to move down one place. Similarly, an item can bk transferred out of the column only from the top, and when this happens all the remaining items move up one place.

16-bitQ-store core plane

Page 12: very speed data processing system - Computer History Museumarchive.computerhistory.org/resources/text/English...a very hgh speed data processing system s5 k 0 'for Commerce, Industry,

Current Operadon Nesting Store Most arithmetic and other operations in the KDF9 take place in the current operation nesting store, the operands and the result being stored in the top few locations. The exact number of locations actively concerned varies from one to four depmding on the operation being performed.

An important feature of the store is that arithmetic operations and certain functions such as duplication of an operand, reversing the relative position of operands, and erasure, do not require reference to the main store. This feature increases the speed of computation and effects a considerable economy in program instructions. Furthermore, by use of the modular instruction code, instructions can be shorter than hitherto.

NESTING STORES PERMIT EASIER PROGRAMMING, FASTER COMPUTING, AND SHORTER INSTRUCTIONS

Subrwtine Jump N d n g Swre In addition to the current operation nesting store, there is a nesting store of sixteen 18-bit registers used in entering and leaving subroutines. On entering a sub- routine the next instruction location is stored in the top register, the previous contents moving down. On leaving the subroutine, control is transferred to the instruction whose address is in the top location, and the other contents move up. In this way closed subroutines of up to sixteenth order can be handled very simply and subroutines of even higher orders can be conveniently accommodated.

ADDRESS MODIFICATION ANQ COUNTLNG (Q-STORES) There are 15high-speed addressable registers called Q-stores which may be used for address modifica- tions and counting, in which case they hold three separate 16-bit numbers : a counter, an increment, and a modifier. Each main store transfer instruction specifies a Q-store, the modifier part of which is added to the main address before execution ; the instruction may also specify that after the transfer the increment is added to the modifier and the counter is decreased by one. The three parts of a Q-store are separately addressable and may be used as temporary stores or as accumulators ; thus the assembly of Q-stores may be looked upon as 15 48-bit high-speed stores or accumulators, or as 45 16-bit high-speed stores or accumulators. Data are transferred into a Q-store from the top cell of the nesting store and vice versa.

Q-STORES PERMIT EASIER PROG-G, FASTER COMPUTING, AND SHORTER INSTRUCTIONS

Q-STORE : S T O R A G E OF 4 8 - B I T S (TOTAL NUMBER OF Q-STORES : 15)

C O U N T E R I I N C R E M E N TI i I Y Y I

ADDRESSABLE 16-BIT ADDRESSABLE 16-BIT ADDRESSABLE 16-BIT HIGH-SPEED STORE HIGH-SPEED STORE HIGH-SPEED STORE I OR ACCUMULATOR OR ACCUMULATOR OR ACCUMULATOR

ADDRESSABLE BIT HIGH-SPEED STORE OR ACCUMULATOR

Page 13: very speed data processing system - Computer History Museumarchive.computerhistory.org/resources/text/English...a very hgh speed data processing system s5 k 0 'for Commerce, Industry,

The breakdown of the

IT M A Y BE

USED A S

EIGHT 6-BIT ALPHA- NUMERIC CHARACTERS

ONE 48-BIT FIXED-POINT NUMBER

TWO HALF-LENGTH FIXED-POINT NUMBERS

HALF DOUBLE-LENGTH FIXED-POINT NUMBER

ONE 48-BIT FLOATING-POINT NUMBER

TWO HALF-LENGTH FLOATING-POINT NUMBERS

*HALF DOUBLE-LENGTH FLOATING-POINT NUMBER

Page 14: very speed data processing system - Computer History Museumarchive.computerhistory.org/resources/text/English...a very hgh speed data processing system s5 k 0 'for Commerce, Industry,

6 BiTS 6 6 BITS 6 6 BITS1 1 1 TH-7 I I

V""E (C-128)OF NUMBER: f x 2

39-BIT FRACTION (f)

(C-128)VALUE OF NUMBER: f x 2

15-FRACTION (f)

(C-128)VALUE OF NUMBER : f x 2

.-

%--IT FRACTION ( f )

a-Ol l 39-BIT CHAR (c) FRACTION If \

*CHARACTERISTIC OF LEAST SIGNIFICANT HALF IS 39 LESS THAN THAT OF MOST SIGNIFICANT HALF.

Page 15: very speed data processing system - Computer History Museumarchive.computerhistory.org/resources/text/English...a very hgh speed data processing system s5 k 0 'for Commerce, Industry,

Programming

The KDF9 instruction code achieves remarkable economy in program storage space and

clearly reflects the wealth of 'English Electric ' experience in computer programming and

automatic programming schemes. Programming is exceptionally easy and the code facilitates

translation of programs written in any new or existing psuedocode into an efficient machine

program. This is a valuable feature of considerable importance to potential users who have

large capital assets in existing programs compiled for earlier and slower machines.

MODULAR INSTRUCTION CODE. The instruction format is based on the novel concept of a modular or variable-length

instruction word, each instruction occupying only sufficient digits to specify the operation

concerned. To achieve this the word is divided into six " syllables " of eight bits each, and

an instruction may occupy one, two, or three syllables.

Instructions to fetch or store numbers in the main store use three syllables and this is

sufficient to specify the address of any location, and how the address is to be modified.

In addition to main store transfers the instructions cover conditional or unconditional

jump operations.

The majority of instructions, however, need only one syllable and cover the operations in

the nesting store. They include all arithmetic and logical operations and also certain special

operations for manipulating and rearranging the top four items of data in the nesting store.

This results in striking economy of storage space for a given program.

The two-syllable instructions specify various ranges of operations including peripheral

transfers, transfers to or from Q-stores, shifting and manipulating data stored more than

one item to a word, and indirect addressing of the main store (including half-length transfers).

Page 16: very speed data processing system - Computer History Museumarchive.computerhistory.org/resources/text/English...a very hgh speed data processing system s5 k 0 'for Commerce, Industry,

RANGE OF INSTRUCTIONS

Main Store Transfers Transfers between the main store and nesting store may be on full-

length or half-length words. In half-length warking every half-word of the main store is

separately addressable. Half-length working is particularly valuable when low precision

working is acceptable as it doubles the effective size of the main store.

Arithmetic Operations A comprehensive list of arithmetic operations is available ;

these include a variety of single, double, and mixed-length operations on both fixed and

floating-point numbers, and a generous supply of manipulative and discrimination

instructions. There are facilities for conversion between binary and any single-word mixed-

radix representation in 6-bit characters that are particularly valuable for data processing

work.

Peripheral Transfers A transfer of data between the main store and a peripheral device is

initiated by an instruction which merely specifies the main store location, the identity of the

peripheral unit, the number of words to be transferred, and the operation to be performed.

The transfer then takes place autonomously and proceeds in parallel with internal or other

external operations.

INTERLOCICS There are full interlocks on the instructions which transfer data so that the programmer has

no problem in timing such transfers. The appropriate area of the main store remains

inaccessible to any other operation until completion of the transfer and all automatic checks

within a block. On a computer not equipped with time-sharing facilities the user may have

several peripheral devices working simultaneously, each with its own area of core store

securely interlocked. The interlocks allow all other operations in the computer to proceed,

provided only that they do not make reference to the interlocked area. If so, then a hold-up

occurs until the peripheral transfer is cleared.

If the computer is equipped with time-sharing facilities (described in next section), instead of

a hold-up occurring an alternative lower priority program will be entered and obeyed while

data transfers are taking place. Thus time is utilised which, on a conventional machine,

would be lost during interlock periods.

Page 17: very speed data processing system - Computer History Museumarchive.computerhistory.org/resources/text/English...a very hgh speed data processing system s5 k 0 'for Commerce, Industry,

Control Systems

The KDF9 system provides for a combination of units which will meet the widest possible range of requirements. In consequence, alternative main control units are available : Conventional, in which operations proceed in normal sequence (apart from the autonomous operation of peripheral transfers already described), Time-sharing, and Time-Smoothing Advance Control.

The system can handle up to four completely independent programs at one time, each program being given a priority rating by the operator. Each program is allocated its own required area of the main store and set of peripheral devices, the program designated as the highest priority proceeding uninterrupted until some delaying condition arises. For example, data may be required from a magnetic tape unit and no further work can be carried out until the required block has been read into the main store. In these circumstances interrup- tion takes place and work on this program lapses. A second-priority program is then started and continues until it, too, is delayed or the first-priority program is ready to continue.

Comprehensive automatic interlocks are provided to prevent any possibility of inadvertent interference between programs, and both commercial and scientific programs can be performed concurrently if desired. If a program attempts to use equipment not allocated to it, it is interrupted, a failure message printed, and control is transferred to another program.

The allocation of equipment to programs, program interruptions, and failure indications, is handled by a stored supervisory routine.

The diagram opposite depicts a typical time-sharing system, but showing only a " token " set of peripheral devices per program. It also indicates that the four areas of core storage are varied in size to suit the actual program length, thus allowing considerable flexibility in the storage of programs.

TIME-SMOOTHmTG ADVANCE CONTROL A further available feature is the time-smoothing advance control, the function of which is to level out peak demands on the main storage and on arithmetic operations. By anticipating future requirements for data from the main store, and by allowing other operations to proceed during the transfer of data into the mam store, the already high speed of the computer may be increased by a factor of up to 35%. No additional programming is required ; programs written for any alternative control unit can be used without alteration.

Page 18: very speed data processing system - Computer History Museumarchive.computerhistory.org/resources/text/English...a very hgh speed data processing system s5 k 0 'for Commerce, Industry,

The KDFS Time-sharing system (DIRECTOR IN CONTROL - SWITCHING BETWEEN PRIORITIES)

I INPUT / OUTPUT I INPUT / OUTPUT I I INPUT / OUTPUT I I INPUT / OUTPUT I II

PAPER r n u u I PRINT TAPE I OUT CARDS I ni IT

MAGNETIC MAGNETIC TI -- - - -E

CORE STORE

EACH PROGRAM

OCCUPIES EXACTLY

THE AMOUNT OF

CORE STORAGE

ACCESSIBLE UNITS ADDRESSING SWITCHED

USED &I DMCCTOI BY DIRECTOR

I

Q = Q-STORE (ADDRESS MODIFICATION AND COUNTING)

NS = CURRENT OPERATIONS NESTING STORE

SINS = SUBROUTINE JUMP NESTING STORE

Page 19: very speed data processing system - Computer History Museumarchive.computerhistory.org/resources/text/English...a very hgh speed data processing system s5 k 0 'for Commerce, Industry,

The KDFS Time-sharing system (PROGRAM OPERATING)

I , INPUT / OUTPUT I I INPUT / OUTPUT I I INPUT / OUTPUT INPUT / OUTPUTl I I I

MAGNETIC TAPE

CORE STORE

EACH PROGRAM

OCCUPIES EXACTLY

CORE STORAGE

ACCESSIBLE UNITS CONTROL UNlT ADDRESSING SWITCHED I ARITHMETIC UNlT

JSED BY DIRECTOR BY DIRECTOR - WORKING REGISTERS

Q = Q-STORE (ADDRESS MODIFICATION AND COUNTING)

NS = CURRENT OPERATIONS NESTING STORE

SJNS = SUBROUTINE JUMP NESTING STORE

Page 20: very speed data processing system - Computer History Museumarchive.computerhistory.org/resources/text/English...a very hgh speed data processing system s5 k 0 'for Commerce, Industry,

Operating Speed

Mixd Rirdix Conversion (to or from b-1 (dependent an word make-up)

(a) Facilities for Wuft ion modifscation and wunliq are parCfcularly good.

Page 21: very speed data processing system - Computer History Museumarchive.computerhistory.org/resources/text/English...a very hgh speed data processing system s5 k 0 'for Commerce, Industry,

COMMERCIAL OPERATIONS. A sorting program that occupies only 35 words has been made for KDF9. In an example with 2,000 items of variable length, whose average length is four words per item, the time taken to sort into numerical sequence from random order by the standard two-way merge process is less than three seconds.

KDF9 is very suitable for production and stock control problems, and its high operating speed enables management in large and complex organisations to exercise the same tight control that hitherto was possible only in the " one man " business.

In stock control problems a forecast* for next week's usage of a particular item may be made from smoothed historical data of actual usage and its trend, by application of the formula :

where : f = week's forecast usage. Y = last week's actual usage. b = trend of usage. t = subscript representing applicable week number. a = smoothing constant.

'The evaluation is done by a KDF9 routine which uses only six words of storage space and takes only 155 microseconds.

TYPICAL COMPUTATZON TIMES The calculation of the expression :

where a, b, c, and d are fixed or floating-point numbers stored in successive store locations, takes less than 180 microseconds by a program occupying only five words (including fetches). This time is reduced if the time-smoothing advance control is used.

The diagram on page 18 shows the step-by-step nesting store functions and locations occupied during the calculation of the above expression. The variable-length instruction code is also depicted, from which it will be noted that the fourth three-syllable " fetch " actually runs from one word to the next.

A simultaneous linear equations routine occupies only 36 words of instructions and, using the method of successive elimination with interchanges, solves 100 equations in 15 seconds without advance control. With advance control this speed may be increased by a factor of up to 35 %.

* Ref: "Statistical Forecasting for Inventory Control" -R. G. Brown

Page 22: very speed data processing system - Computer History Museumarchive.computerhistory.org/resources/text/English...a very hgh speed data processing system s5 k 0 'for Commerce, Industry,

Use of KDF9 instruction code to evaluate a(a3b2+1)

OPERATION

FETCH . .

FETCH . .

DOUBLE DUP.

DUPLICATE . . MULTIPLY . .

MULTIPLY . .

DUPLICATE . .

MULTIPLY . .

ADD

REVERSE . .

FETCH d . .

FETCH c . .

MULTIPLY . .

DUPLICATE . .

N1

b

a

a

a

a2

a2 b

a2 b

a4 b2

a4 b2+a

b

d

C

cd

:d

a

a' b

a

b

NUM

b NUM -

d b NUM

b NUM -

cd b NUM

MULTIPLY . . ;? d 2 b NUM -

EXAMPLE O F DUPLICATE . . -2 d ? 32 d2 b N'UM

VARIABLE 4 D D . . 2c2d2 b NUM -

LENGTH 4 D D . . 3ENOM NUM

lNSTRUCTlON NUMDIVIDE . . )ENOM

STORE . . -

TOTAL WORDS (W): 5 TOTAL SYLLABLES: 30

Main Store Instruction R Nesting Store Instruction

Page 23: very speed data processing system - Computer History Museumarchive.computerhistory.org/resources/text/English...a very hgh speed data processing system s5 k 0 'for Commerce, Industry,

Peripheral Devices

A very comprehensive range of peripheral devices is available for use with KDF9. In addition to the 'English Electric' high speed paper tape reader, the devices include the high performance magnetic tape units, high speed line printers, and card readers and card punches of the 'English Electric ' KDPlO Data Processing System. The KDPlO range of off-line transcribers may also be used with KDF9.

ALL THESE DEVICES HAVE BEEN TESTED AND PROVEN M SERYICE

There is no practical limit to the total number of peripheral devices which can be connected into the system, and any number can operate simultaneously, subject to a maximum theoretical total instantaneous transfer rate of 1.33 million characters per second.

The channels of communication with the peripheral equipment have been designed to allow easy addition of existing devices or any further devices which may be developed in the future.

Page 24: very speed data processing system - Computer History Museumarchive.computerhistory.org/resources/text/English...a very hgh speed data processing system s5 k 0 'for Commerce, Industry,

MAGNETIC TAPE UNlT!3

The magnetic tape units-already fully proven in the KDPlO system-are designed to provide the maximum accuracy and reliability in operation, and special attention has been given to safeguards against flaws or errors. Each unit is fully transistarised and is used for reading, writing, and erasing characters on magnetic tape in response to applied control signals. Reading is possible in the forward or reverse direction ; writing and erasing are possible only in the forward direction.

The magnetic tape is $ in. wide and contains 16 recording channels. Information is recorded simultaneously across eight channels and duplicated on the other eight channels. In each group six channels are for information, one for parity, and one for timing.

When data are read from one set of channels the signal received is merged with the signal from the adjacent set and a correct output is obtained even if a bit from one channel is completely missing. This ensures protection against " drop-outs " due to dust particles or small local imperfections in the tape. The accuracy of information read from the tape is further ensured by means of a parity check.

The high performance of the tape unit is founded upon a recording density of 333-3 characters per inch, a startlstop time of 2.5 milliseconds, and very short inter-block gaps of 0.5 in average.

I I The magnetic tape units have a

reading or writing rate of 33,333

9characters per second. Tape units of similar characteristics but oper- ating at the rate of 66,666 characters per second will shortly be available. The increase in transfer rate is achieved without sacrifice of start/ stop time or interblock gap length.

Page 25: very speed data processing system - Computer History Museumarchive.computerhistory.org/resources/text/English...a very hgh speed data processing system s5 k 0 'for Commerce, Industry,

KDF9 is equipped with a medium-sad peper tape pun& which n a d y perforatesone-in& taw in an eight-channel even-parity binmy code. In regponse to multiwimel& h p d m the pwch fufl perforates papa apea the rate of 110 charmrs per second, An dght-ineb diunclcr d h o ~ d s1.W fen of tpp. By suitable output subroutine, tspmay be punch~dIn any code on any sfthe standrud wid& of tape.

The illustration shows the location of the tape-reel and containment of the tape in the magazine.

Page 26: very speed data processing system - Computer History Museumarchive.computerhistory.org/resources/text/English...a very hgh speed data processing system s5 k 0 'for Commerce, Industry,

HIGH SPEED ON-LTNE PRINTER The on-line printer is electrically connected to the computer and prepares output documents at high speed. The printing format is controlled by the computer, either directly or through a punched tape loop in the printer unit. The printer is fully transistorised.

The input rate and the line printing rate are determined by the computer. The printer output is 600 lines per minute, printing single-spaced copy on continuous edge-perforated blank or pre-printed forms with a maximum of 120 characters per line. The vertical align- ment within a horizontal line is better than plus or minus .008 in. per single copy.

The paper stock is single or multiple sheet fanfold and may be from 3 in. to 22 in. wide and up to 17 in. sheet length, providing the length is a multiple of the standard 4 in. sprocket- hole distance. The printer prints one original and three carbons.

A total of 51printed characters are used, comprising 26 English capitals, 10 decimal numbers, and 15 punctuation marks and symbols as under :

, Comma - Minus

; Semi-colon ( Open bracket

: Colon ) Close bracket

. Full stop / Solidus

' Apostrophe & Ampersand

" Quotes & Pound

* Asterisk % Per cent

$$ Number

The printer roller consists of 120 axially mounted rings each having 51 printed characters around its periphery.

Page 27: very speed data processing system - Computer History Museumarchive.computerhistory.org/resources/text/English...a very hgh speed data processing system s5 k 0 'for Commerce, Industry,

INTERRUPTION TYFEWRITER The interruption typewriter is an on-line device, similar to an electric typewriter, which prints on paper from data received directly from the computer store. It is mainly used for two-way communication between the computer and the operator. The interruption type- writer can be used as a monitor printer printing at a rate of ten characters per second, and as a keyboard input in conjunction with an attached edge-punched card reader for controlling the course of a computation. An integral punch produces seven-hole punched paper tape simultaneously with the printed output.

PIJNQIED CARD EQUIPMENT Card readers and punches are available, operating at 400 and 150 cards per minute respectively. The equipments incorporate comprehensive checking arrangements. In reading, each row of a card is read twice and the readings compared ; in punching, each row is read immediately after punching and compared with the data intended to be punched.

Page 28: very speed data processing system - Computer History Museumarchive.computerhistory.org/resources/text/English...a very hgh speed data processing system s5 k 0 'for Commerce, Industry,

Ofl-line Operat ions As the main computer represents a substantial part of the capital cost of an installation, the use of off-line equipment can be of great economic importance. The transfer of input and output operations to off-line working using magnetic tape releases the computer from major time-consuming operations, and ensures that the fullest economic use of high operating speed is obtained. This is particularly important if time-sharing facilities are not available.

If desired, printing may be carried out at a distance from the main computer. For example, printing equipment may be installed at a centre many miles from the computer and the magnetic tape reels containing the output sent by post from the computing centre to the printing centre. One reel of magnetic tape is capable of producing up to six hours' continuous printed output, representing 220,000 to 320,000 lines of print.

The tried and tested off-line transcribers developed for KDPlO may be used with KDF9 ; these include transcription from punched cards to magnetic tape and vice versa, and from magnetic tape to the printers described in the following paragraphs. All transcribers have comprehensive editing facilities.

HIGH SPEED OPFJ.JNE PRlNTER. A printer similar to the on-line printer may be used for simultaneous off-line operation, in which case it accepts coded data from magnetic tape on a separate magnetic tape unit. The printing format is controlled by a plug-board or by a punched tape loop in the printer unit. Operation in the " select mode " enables up to 15 different types of print format to be produced from the same reel of magnetic tape.

The input is accepted from the magnetic tape unit at the rate of 33,333 characters per second, in bursts of one line. The printing output rates and alignment are the same as for the on-line printer.

Because of the asynchronous nature of the operation, numeric data are printed at the rate of 900 lines per minute and the output speed will always lie between 600 and 900 1.p.m. depending upon the ratio of alphabetical to numerical content of the data.

The off-line printer uses the same 51 printed characters as the on-line printer and also the same paper stock.

An off-line Xeronic printer receiving its input from magnetic tape may also be used. The printer prints up to 3,000 lines per minute printing either two columns (128 characters per column line) or four columns (64 characters per column line), the line spacing being six lines per inch. A standard form layout, selected from up to 20 alternatives, can be printed on blank paper simultaneously with the data.

Page 29: very speed data processing system - Computer History Museumarchive.computerhistory.org/resources/text/English...a very hgh speed data processing system s5 k 0 'for Commerce, Industry,

User Services

KDF9 Is e~tremolyeasy to p m p m , but nevertheless a wide nmge of autmntiic pmgmm-min$ BFbomes is am.&bk, Zaparticnlar, them an:ofEdeat versions of ALGOL and othttr &%G and m# autcrcodm w U will appeal to atl who wisb ta w propam written and mted an othe* mstchme~,Such progt'gnts may be BCCeptled dimctfy by KDFO. In addition, an t~t.t&:mctypowerful matrix sFhtffu: is being piqatwl MI that all problems whit& ~ 8 nbe oxpFesssd in mmtsrix natation ean be progtammed aa mily rut the matrix cqu8tistllg em ttc writtea

Page 30: very speed data processing system - Computer History Museumarchive.computerhistory.org/resources/text/English...a very hgh speed data processing system s5 k 0 'for Commerce, Industry,
Page 31: very speed data processing system - Computer History Museumarchive.computerhistory.org/resources/text/English...a very hgh speed data processing system s5 k 0 'for Commerce, Industry,

CONSTRUCTIONAL

Unit Height Length Depth Weight

Control desk with interruption typewriter

2 ft. 6 in. (076 m.)

6 ft. 0 in. (1.83 m.)

2 ft. 9 in. (0.84 m.)

300 lb. (136 kg.)

*Main store and input/output control unit

*Arithmetic and main control unit

Power supply unit

5 ft. 9 in. (1.75 m.) 5 ft. 9 in. (1.75 m.) 5 ft. 9 in. (1.75 m.)

18 ft. 4 in. (5.59m.) 18 ft. 4 in. (5.59 m.) 7 ft. 4 in. (2-24 in.)

1 ft. 7 in. (0.48 m.) 1 ft. 7 in. (0.48m.) 1 ft. 7 in. 10-48 m.)

3,500 lb. (1,587 kg.) 3,500 lb. (1,587 kg.) 3,000 lb.

(1,360 kg5 Paper tape reader 4 ft. 9 in. 2 ft. 7 in. I ft. 8 in. 280 lb.

(1.45 m.) (0.79 m.) (0.51 m.) (127 kg.) Paper tape punch On control desk or separate table 27 lb.

Card reader 3 ft. 9 in. 4 ft. 0 in. 2 ft. 6 in. (12 kg.) 500 lb.

(1.14 m.) (1.22 m.) (0.76 m.) (227 kg.) Card punch 3 ft. 9 in. 4 ft. 0 in. 2 ft. 6in. 400 l b

(1.14 m.) (1.22 m.) (0.76 m.) (181 kg.) Line printer 3 ft. 9 in. 3 ft. 9 in. 2 ft. 6 in. 500 lb.

(1.14 m.) (1.14 m.) (0.76 m.) (227 kg.) Printer control 5 ft. 9 in. 7 ft. 4 in. 1 ft. 7 in. 600 lb.

(1.75 m.) (2.24 m.) (0.48 m.) (272 kg.) Magnetic Tape Unit 5 ft. 9 in. 3 ft. 8 in. 1 ft. 7 in. 900 lb.

(1.75 m.) (1.12 m.) (0.48 m.) (408 kg.)

POWER REQUIREMENTS

All units 200 to 250V 50 c/s single-phase Control desk with interruption typewriter -

*Main stores with input/output control unit 2.4 kW *Arithmetic and main control unit 1.7 kW Power supply unit 1.6 kW Magnetic Tape Unit 1.5 kW Paper tape reader 0.5 kW Paper tape punch 0.1 kW Card reader 1.9 kW Card punch 1.5 kW Line printer 3.2 kW

* Figures relate to a typical installation having a 16,384 word main store, time-sharing control, and a total of eight input/output devices

Page 32: very speed data processing system - Computer History Museumarchive.computerhistory.org/resources/text/English...a very hgh speed data processing system s5 k 0 'for Commerce, Industry,

After Sales Service

'English Electric' were among the pioneers of computer maintenance service in this country, and are able to remove completely from the user the onus of responsibility for the main- tenance of a system. Proposals can be put forward by our Service Department to cover combinations of the following features :

1. Comprehensive service by resident engineers covering one, two, or three shifts of opera- tion for up to seven days per week, or non-standard timetable to cover specific demanded cases.

2. The provision of a backing service of skilled electronic and mechanical engineers to ensure, by regular inspection, the maintenance of the equipment to a highly efficient standard, particularly the mechanical parts of peripheral equipment.

3. The provision of reserve items of peripheral equipment, held at base, such as tape stations and line printers, thus covering cases of an accident or major breakdown at a Customer's site.

4. The provision of initial spares and test gear either as an initial capital investment, or as part of an annual service charge.

5. A comprehensive annual service charge or, alternatively an annual service charge based on replacement at the Customer's expense of certain parts subject to wear (magnetic tape unit recording heads, print rolls and hammers), should this be deemed necessary subsequent to the first three shift years of operation.

6 . In circumstances in which the Customer desires to provide his own first line maintenance, the provision of services as detailed in (2) above.

The KDF9 Computer Service can be as comprehensive as you

wish -and your expenditure can be accurately budgeted.

Note: The Company's policy is one of continuous development and improvement of' its products and, therefore, the right is reserved to supply products which may differ slightly from those described and illustrated in this publication.

The designations DEUCE, KDPIO, KDF9, DATAPAC and LACE are trade-marks of 'English Electric'

Page 33: very speed data processing system - Computer History Museumarchive.computerhistory.org/resources/text/English...a very hgh speed data processing system s5 k 0 'for Commerce, Industry,

The ' E N G L I S H ELECTRIC' Data Processing System The preceding pages will have shown that KDF9 is the most advanced digital computing system yet devised. Nevertheless, it is worthwhile repeating some of the more important features, and these are listed below :

Extremely fast operating speed

Simple and economical programming

Ease of translation from any psuedocode to an efficient machine program

Time-sharing between up to four independent programs

Comprehensive range of fast and flexible peripheral devices

Simultaneous operation of peripheral transfers up to a maximum theoretical total transfer rate of 1.33 million 6-bit characters per second

Wide range of strictly designed arithmetic operations, single, double, and half-length, fixed and floating-point

Main store size and number of peripheral units easily expansible to suit work load

Design and construction techniques based on maximum reliability and ease of maintenance

For further infornlation write to .

'ENGLISH ELECTRIC' DATA PROCESSING AND CONTROL SYSTEMS DIVISION,

KIDSGROVE, STOKEON-TRENT S el. : Kidsgrove 2141

London Sales Office and London Computing Service,

Queens IEouse, Kingsway, London, W.C.2 Tel. :Covent Garden 1234

Page 34: very speed data processing system - Computer History Museumarchive.computerhistory.org/resources/text/English...a very hgh speed data processing system s5 k 0 'for Commerce, Industry,

Science and Engineering Research Council

Rutherford Appleton Laboratory CHILTON, DIDCOT, OXON. OX11 OQX . Tel: Abingdon (0235) 21900 Telex: 83159

I Prof F R A Hopgood

A t l a s C e n t r e

E x t =7 $S22-

Data Processing and Control Sys~em Division

KIDSGROVE, STOKE-ON-TRENT

London Sales Ofice and London Computing Service :

QUEENS HOUSE, KINGSWAY, LONDON W.C.2